Nearly 30 R&D organizations registered for last week’s Alchemite™ User Group meeting, spanning sectors including chemicals, consumer products, food, plastics, alloys, construction chemicals, advanced materials, and life sciences. The meeting shared experiences of successful applications of machine learning, updated users on the latest developments in the software, and provided a glimpse into future developments in machine learning and agentic AI – inviting user collaboration in ensuring these advances meet their key requirements.
Users share success
The highlights were three case study presentations from user organizations showcasing applications of Alchemite™ in the chemicals and materials sectors. Full details are available to meeting participants, but key learnings were:
- A leading chemical company used Alchemite™ for adaptive design of experiments, enabling their team to home-in faster on formulation development targets. The presenter emphasized how well Alchemite™ fitted with their existing iterative development workflow and discussed future plans, including extending use to big data on materials, formulations, and product tests.
- A long-standing Alchemite™ user organization assessed the new SMILES extension functionality as a tool for enabling chemically-informed machine learning, finding that it performed well in the prediction of key solubility parameters. This organization is now looking into building a screening tool for reformulators that exploits the SMILES capability.
- Another user described two separate use case examples for Alchemite™. The first focused on rheology properties, predicting a number of formulations that met target criteria, one of which has been tested successfully. The second targeted formulations where there was a need to balance two conflicting properties. The ‘calculated columns’ feature was used to code domain knowledge into the ML model. The project recommended three formulations, two of which have showed promise when synthesized and tested at lab scale.
New features showcased
The Intellegens team provided updates on recent developments. New functionality in the Alchemite™ Suite front end included enhanced analytics for DOE, tools to create better models, integration with Excel, use of generative AI, and improved admin capabilities. Key advances in the underlying Alchemite™ method were increased accuracy by building-in targets during model building and improved estimates of uncertainty, especially when extrapolating. There was also discussion of the new SMILES extension, which makes Alchemite™ models aware of chemical structure.
A glimpse of the future
Plans for Alchemite™ Academy were introduced. This new modular training and learning program will help customer organizations to gain maximum benefit from machine learning, data analysis, and adaptive DOE. Product roadmap sessions looked at future developments in data and integrations, sustainability tools, security, and ease-of-use. There was also a live demonstration of ongoing development work in agentic AI.
Intellegens CEO Ben Pellegrini concluded the meeting with an invitation to engage in collaborative opportunities, particularly around training and agentic AI, and for users to continue active conversations, providing the Intellegens team with feedback and ideas.
